Comprehensive Psychiatry 106 (2021) 152223

Contents lists available at ScienceDirect

Comprehensive Psychiatry

journal homepage: www.elsevier.com/locate/comppsych

Cognitive behavioural therapy with exposure and response prevention

in the treatment of obsessive-compulsive disorder: A systematic
review and meta-analysis of randomised controlled trials
Jemma E. Reid a,b,⁎, Keith R. Laws a, Lynne Drummond c, Matteo Vismara d, Benedetta Grancini d,
Davis Mpavaenda a,b, Naomi A. Fineberg a,b,e
a
University of Hertfordshire, Hatfield, Hertfordshire, UK
b
Hertfordshire Partnership University NHS Foundation Trust, Hertfordshire, UK
c
South West London and St George's NHS Trust, UK
d
University of Milan, Department of Biomedical and Clinical Sciences Luigi Sacco, Milano, Italy
e
University of Cambridge School of Clinical Medicine, Cambridge, UK

a r t i c l e i n f o a b s t r a c t

Background: Cognitive behavioural therapy (CBT), incorporating exposure and response prevention (ERP) is
widely recognised as the psychological treatment of choice for obsessive-compulsive disorder (OCD). Uncer-
Keywords:
tainty remains however about the magnitude of the effect of CBT with ERP and the impact of moderating factors
Cognitive behavioural therapy in patients with OCD.
Exposure and response prevention Method: This systematic review and meta-analysis assessed randomised-controlled trials of CBT with ERP in pa-
researcher allegiance tients of all ages with OCD. The study was preregistered in PROSPERO (CRD42019122311). The primary outcome
Meta-analysis was end-of-trial OCD symptom scores. The moderating effects of patient-related and study-related factors in-
Obsessive-compulsive disorder cluding type of control intervention and risk of bias were examined. Additional exploratory analyses assessed
the effects of treatment fidelity and impact of researcher allegiance.
Results: Thirty-six studies were included, involving 2020 patients (537 children/adolescents and 1483 adults)
with 1005 assigned to CBT with ERP and 1015 to control conditions. When compared against all control condi-
tions, a large pooled effect size (ES) emerged in favour of CBT with ERP (g = 0.74: 95% CI = 0.51 to 0.97 k =
36), which appeared to diminish with increasing age. While CBT with ERP was more effective than psychological
placebo (g = 1.13 95% CI 0.71 to 1.55, k = 10), it was no more effective than other active forms of psychological
therapy (g = −0.05: 95% CI −0.27 to 0.16, k = 8). Similarly, whereas CBT with ERP was significantly superior
when compared to all forms of pharmacological treatment (g = 0.36: 95% CI 0.7 to 0.64, k = 7), the effect became
marginal when compared with adequate dosages of pharmacotherapy for OCD (g = 0.32: 95% CI −0.00 to 0.64,
k = 6).A minority of studies (k = 8) were deemed to be at low risk of bias. Moreover, three quarters of studies
(k = 28) demonstrated suspected researcher allegiance and these studies reported a large ES (g = 0.95: 95% CI
0.69 to 1.2), while those without suspected researcher allegiance (k = 8) indicated that CBT with ERP was not
efficacious (g = 0.02: 95% CI −0.29 to 0.33).
Conclusions: A large effect size was found for CBT with ERP in reducing the symptoms of OCD, but depends upon
the choice of comparator control. This meta-analysis also highlights concerns about the methodological rigor and
reporting of published studies of CBT with ERP in OCD. In particular, efficacy was strongly linked to researcher
allegiance and this requires further future investigation.
© 2021 Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

⁎ Corresponding author at: Hertfordshire Partnership University NHS Foundation Trust, Rosanne House, Parkway, Welwyn Garden City, Hertfordshire AL8 6HG, UK.
E-mail address: jemma.reid1@nhs.net (J.E. Reid).

https://doi.org/10.1016/j.comppsych.2021.152223
0010-440X/© 2021 Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
J.E. Reid, K.R. Laws, L. Drummond et al. Comprehensive Psychiatry 106 (2021) 152223

3.1. Primary outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3.2. Secondary outcomes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2.1. Sub-groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.2.2. Moderators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.3. Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.4. Exploratory outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.4.1. Treatment fidelity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.4.2. Researcher allegiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3.5. Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.1. Publication bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.2. Moderator and exploratory analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.3. Control type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.4. Researcher allegiance bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.5. Adults vs children . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.6. Group vs individual therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.7. Other moderator analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.8. Risk of bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1. Introduction
Obsessive Compulsive Disorder (OCD) is a highly debilitating and
disabling illness, associated with significant impairment both of the
quality of life of the affected individual and on a wider societal scale in
terms of loss of productivity and functioning (Hollander et al. [1]).
OCD is relatively common with a 12-month prevalence of approxi-
mately 1.2% (DSM-5) [2]. The illness usually emerges in childhood or
early adulthood and runs a chronic, relapsing course (Fineberg et al.
[3]). Detection of OCD frequently occurs late and many patients experi-
ence untreated illness for a significant length of time before receiving
treatment (Dell'Osso et al. [4]). Increasingly, evidence suggests that a
longer duration of untreated illness leads to poorer outcomes and prog-
nosis (Fineberg et al.) [5]. Therefore, it is of paramount importance that
patients with OCD receive appropriate treatment in a timely manner to
reduce suffering and improve functioning.
Recommended treatments for OCD include psychological therapy
with cognitive behavioural therapy (CBT) involving exposure and re-
sponse prevention (ERP) (ERP is a therapy in which patients are taught
to confront and tolerate conditions that provoke obsessions and com-
pulsions and resist acting on them) or pharmacotherapy with a selective
serotonin reuptake inhibitor (SSRI) or the serotonergic tricyclic clomip-
ramine. As SSRI in OCD shows a positive dose-response relationship [6]
the highest available dosages are recommended [7].The influential 2005
NICE guidelines (CG31) [8], which were based on a meta-analysis of
existing trial data, advocate the use of low intensity psychological treat-
ments (including ERP) for adult patients with mild symptoms of OCD.
Monotherapy with either more intensive CBT (including ERP) or an
SSRI is recommended for patients with moderate symptoms or patients
with mild illness who cannot tolerate low-intensity psychological treat-
ment, whereas combination therapy (SSRI and CBT with ERP) is recom-
mended for patients with more severe or resistant illness. In the case of
children and young people with OCD, CBT is prioritised over pharmaco-
therapy, to avoid potential adverse effects of medication in this age
group and ERP is cited as the recommended type of CBT [8]. However,
as the original analyses upon which this guidance is based is now
more than 15 years old and as more data has since accrued (NICE
have stated support for a review of the OCD treatment guidelines)
[8,9], it is timely to review the evidence supporting the effectiveness
of CBT involving ERP across the age range in OCD.
A large number of individual studies, varying in quality and size,
have demonstrated that ERP can be an effective treatment for OCD.
These were reviewed in detail in the American Psychiatric Association
practice guidelines [7] which also concluded that CBT primarily based
on behavioural techniques such as ERP has the strongest evidence
base for efficacy. In contrast, the more recently updated British Associa-
tion of Psychopharmacology guidance cites evidence for ERP monother-
apy, cognitive therapy as a monotherapy and a combination of the two
as being effective [10]. Indeed, both documents acknowledge that,
based on the available evidence, we cannot yet determine which ele-
ments of CBT are most responsible for its success. What is, however,
clear is that determining the precise type of CBT delivered from reading
the descriptions given in many of the published treatment trials can be
difficult. It is also evident from the variability within published studies
and subsequent meta-analyses (see below) that models and standards
vary. Of note, a recent small study in adults (Fineberg et al. [11]) that
compared Sertraline monotherapy, CBT with ERP monotherapy deliv-
ered strictly according to a manualised protocol, and combination (Ser-
traline and CBT with ERP) therapy, found disappointing results for CBT
with ERP. Whereas combination therapy was the most efficacious treat-
ment option at 16 weeks, the advantage was not sustained and sertra-
line monotherapy was both the most efficacious and cost effective
option at the 52 week endpoint.
Previous meta-analytic evidence has largely focused on CBT rather
than specifically on CBT with the ERP. For example, a recent meta-
analysis of pediatric OCD by Uhre et al. [12], analysed 12 randomised
controlled trials comparing CBT to wait-list, psychological placebo or
pill placebo. Although symptoms (as measured by change in CY-BOCS)
were significantly reduced by CBT (MD: -8.51, 95% CI: −10.82 to
−6.18), all trials included in the analyses were deemed to be at high
risk of bias and the certainty of evidence was graded as ‘low’ or ‘very
low’. It is important to note that some authors challenged the findings
on methodological grounds [13], generating a debate in the journal's
pages [14,15]. A network meta-analysis of CBT and pharmacotherapy
for adults with OCD by Skapinakis et al. [16,17] was unable to find a
clear advantage of one form of treatment over the other. The study iden-
tified that most of the patients within the studies who were allocated to
CBT were also taking pharmacological treatment, suggesting that these
were in fact trials of combination treatment, and further highlighting
difficulties in interpreting the results from studies of CBT in OCD
(Skapinakis et al. [18]).
Existing meta-analyses have rarely exclusively focused on ERP stud-
ies, with most analysing ERP in sub-group analyses of multiple interven-
tions. The earliest ERP for OCD meta-analytic finding often cited is that
by Christensen et al. [19] who reported a large effect size (2.34), but
this was derived from a pre-post analysis of just one trial. Almost a

2
J.E. Reid, K.R. Laws, L. Drummond et al.
decade later Abramowitz (1996) [20] meta-analysed a substantial cor-
pus of 24 trials (29 samples) assessing the impact of ERP on OCD in
adults and reported a large effect size of 1.16 for pre-post changes.
Pre-post effect sizes however are likely to provide unreliable and in-
flated effect size estimations because of their lack of a control compari-
son (see Cuijpers et al. [21]). It is also notable that most studies in this
meta-analysis (17/29) had very small samples, with fewer than 10 par-
ticipants, which is also likely to produce less reliable findings. A second
meta-analysis by Abramowitz [22] one year later did examine
randomised controlled trials in adult patient samples for whom OCD
was the primary diagnosis. The analysis included eight comparisons be-
tween versions of ERP and other psychological interventions. The study
reported a large effect (using Cohen's d) favouring ERP when relaxation
was used as a psychological control treatment (d = 1.18; 2 studies),
whereas when ERP was compared to cognitive therapy (d = −0.19; 4
studies) or individual components of ERP (i.e. response prevention or
exposure only) no significant effect of ERP was found (d = 0.59; 2 stud-
ies). All 8 studies except one used self-report outcome measures and in-
volved in total only 137 participants who received ERP and 105 controls.
Around the same time Kobak et al. [23] also reported a large effect size
for ERP (0.99 [0.89 to 1.08]) across 36 studies; however this analysis
pooled data from within (pre-post) changes and end-point between
group changes. Later, another pre-post meta-analysis by Eddy et al.
[24] also reported a large effect size of 1.53 for ERP in 16 studies. As
noted, such analyses inflate effect sizes and a further analysis was con-
ducted comparing ERP with controls, but this included just 2 trials and
indicated an effect size of 1.16.
In a meta-analysis involving both experimental and quasi-
experimental designs, Rosa-Alcázar et al. [25] reported a large pooled
effect size for ERP in 13 samples (1.127, 0.80 to 1.45) and this was not
significantly larger than for cognitive restructuring (CR) alone (1.09)
or ERP plus CR (0.998).
Turning to more recent meta-analyses with some component of ERP
assessed in children, McGuire et al. [26] meta-analysed 8 randomised
controlled trials of individually-delivered ERP tested in children only.
In comparisons with non-active controls conditions (mostly waiting-
list and relaxation therapy), they reported a large effect size (g =
1.52), although this was no larger than for trials using cognitive therapy
(1.41). Öst et al. [27] assessed CY-BOCS changes in pediatric OCD both
for individual and for group formats of ERP. The effect size for ERP
(g = 0.68 (95% CI 0.18–1.18, k = 8) was somewhat smaller than that
of McGuire et al. [26] and smaller than that for Cognitive Therapy
(g = 1.04 (95% CI 0.45–1.63, k = 4) with the effect size for ERP + CT
being even smaller and non-significant (g = 0.35 (95% CI −0.04 to
0.73, k = 18). Ӧst et al. [28] also published a meta-analysis of adult trials
and showed that ERP did not differ in efficacy from CBT at reducing Y-
BOCS scores at end-of-trial (0.07 [95%CI -0.15 to 0.30], k = 7) or at
follow-up (0.07 [95%CI -0.27 to 0.41; k = 4).
Olatunji et al. [29] combined trials of CBT and ERP in both adults
(k = 13) and children (k = 3), and in an assessment of 12 ERP trials, re-
ported a large effect size of 1.35 (CI: 0.96–1.74). However, like McGuire
et al. [26], their analysis excluded trials using an active psychological
control and the majority (10/12) of control arms comprised wait-list
controls. The use of wait-list groups in psychotherapy trials is also asso-
ciated with exaggerated effect sizes (Furukawa et al. [30]), and it is rea-
sonable to interpret the efficacy of CBT in such studies with caution.
Intriguingly, Olatunji et al. [29] were also unable to demonstrate any ef-
fect on outcomes of candidate moderators such as age at onset of symp-
toms, duration of illness, gender, number of CBT sessions or the
presence of co-morbidities. However, they did find that the control
group moderated the effect size, with wait-list control comparisons re-
vealing larger effect sizes than comparison to placebo controls.
In summary, data from individual randomised controlled trials and
existing meta-analyses suggest that CBT with ERP is an effective treat-
ment modality for OCD. Concerns about methodological rigor are how-
ever repeatedly highlighted as a limitation on interpreting the available
3
Comprehensive Psychiatry 106 (2021) 152223
data. In particular, the limitations of previous meta-analyses relate to
the assessment of pre-post effect sizes (Abramowitz [20]; Eddy et al.
[24]; Christensen et al. [19]) or mixing pre-post and end of trial effect
sizes (Kobak et al. [23]); the inclusion of small RCTs (e.g. studies by
Abramowitz et al. [22] were mostly <10 per arm); the exclusion of ac-
tive controls and a focus largely on comparisons with wait-list controls
(Olatjunji et al. [29]; McGuire et al. [26]); a reliance on self-report mea-
sures (Abramowitz et al. [22]); meta-analysing small numbers of ERP
studies (McGuire et al. [26], k = 8; Eddy et al. [24] k = 2; Abramowitz
et al. [22] k = 8; Ӧst et al. [28] k = 7; Ӧst et al. [27], k = 8; Christensen
et al. [19] k = 1), which limits the possibility of examining moderator
variables (see Borenstein et al. [31]); the inclusion of non-
experimental designs (e.g. Jónsson, H., & Hougaard, 2008; Rosa-
Alcázar et al. [25]). Some have examined only children (McGuire et al.
[26]; Ӧst et al. [27]), while others only adults (Abramowitz 1996 [20];
Ӧst et al. [28]). Most have failed to address publication bias and earlier
meta-analyses produced effect size estimates based on fixed effects
models. Few existing meta-analyses have assessed study quality (e.g.
Ӧst et al. [28] used a bespoke measure and then examined this in a lim-
ited way rather than as a moderator) and none appear to have used a
standardised measure risk of bias. It is also notable that, where moder-
ators have been analysed, previous meta-analyses have also found it dif-
ficult to confidently identify treatment or patient factors that predict a
better outcome with ERP (Hezel and Simpson, 2019 [32]). Moreover,
as the focus of previous meta-analyses has been primarily on CBT of
any form, rather than one that specifically incorporates ERP, little clarity
exists about the superiority of CBT with ERP over other forms of CBT for
OCD across the full age range affected. Thus, while CBT with ERP re-
mains the suggested psychological treatment of choice for OCD [6], un-
certainty exists regarding its relative efficacy, the methodological
quality and coverage of previous meta-analyses as well as the extent
to which patient or treatment-related factors might render CBT with
ERP the most suitable option for a particular individual.
2. Aims
This meta-analysis aims to comprehensively evaluate the available
evidence from randomised controlled trials addressing the efficacy of
CBT with ERP as a treatment for adults and children with OCD. There-
fore, the analysis only includes studies of CBT that incorporate ERP.
We also aim to identify whether treatment-related or patient-related
factors impact on the treatment-response, in order to aid clinical
decision-making. As concerns about the methodological quality of CBT
studies have been raised in previous reviews (Olantunji et al. [29],
Skapinakis et al. [16]), we aim to conduct a ‘risk of bias’ quality assess-
ment. As it is evident that the CBT delivered in previous studies has
shown considerable variability in quality, we also incorporate an
assessment of the fidelity of the CBT with ERP delivered within this
meta-analysis. In addition, we assess studies for the presence of re-
searcher allegiance (RA), defined as the researchers' “belief in the supe-
riority of a treatment and in the superior validity of the theory of change
that is associated with the treatment” (Leykin & DeRubeis, 2009) [33].
3. Method
This meta-analysis was pre-registered with the International Pro-
spective Register of Systematic Reviews (PROSPERO: registration num-
ber CRD42019122311) as a systematic review and meta-analysis of
cognitive behavioural therapy for OCD. We subsequently refined our
search criteria to focus exclusively on those published studies that in-
cluded an ERP component within the CBT arm, as this is the form of
CBT usually recommended for OCD [7,8]. We conducted a systematic
search of the literature in accordance with PRISMA guidelines (Moher
et al.) [34]. The electronic databases PubMed, PsychINFO and EMBASE
were searched for eligible studies. We also checked the reference lists
of relevant studies and previous systematic reviews for unidentified
J.E. Reid, K.R. Laws, L. Drummond et al.
studies and searched for registered trials on www.ClinicalTrials.gov and
Google Scholar (http://scholar.google.dk). There was no lower limit
with regards to publication date and searches continued until
April 2020.
An inclusive search strategy was performed using the terms: ‘Cogni-
tive behavioural therapy’ OR ‘CBT’ OR ‘exposure response prevention’
OR ‘ERP’ AND ‘obsessive compulsive disorder’ OR ‘OCD’ generated
2265 articles. The articles were then screened using the following inclu-
sion criteria:
1. Randomised controlled trials in patients with OCD involving CBT
with ERP in at least one treatment arm and a control group (which
could be an alternative (non -ERP) psychological treatment, psycholog-
ical placebo, pharmacological treatment or wait-list).
2. The study employed the Yale-Brown Obsessive Compulsive Scale
(Y-BOCS) (or similar symptom severity scale) as an outcome measure.
3. Full text article published in English.
Abstracts were initially screened for relevance by two study authors
(JR and NF). Papers not meeting these criteria were excluded from the
analysis. Accepted studies were then independently assessed by two
members of the team (JR and MV) to evaluate whether they incorpo-
rated ERP into their treatment. Protocol- disagreements were resolved
by discussion and the involvement of a third assessor (NF). A consensus
was reached in all cases.
Our inclusive search strategy located a large number of studies,
which on closer examination were deemed unsuitable as they did not
have a non-ERP comparator treatment arm within their study design.
These studies were subsequently extracted and excluded from the anal-
ysis. Our aim was to include studies where ERP was fundamental to the
CBT being applied. On analysing the studies, it was apparent that signif-
icant variability emerged in the level of description of the included ERP
components. Therefore, where a published report stated that ERP was
an integral component of the CBT being delivered, we included it within
our analysis.
3.1. Primary outcome
Our primary outcome measure was end-of-trial OCD symptom
scores in the CBT with ERP group versus the control group.
3.2. Secondary outcomes
3.2.1. Sub-groups
Sub-group analyses were performed on the studies stratified on the
basis of type of control: Three of the authors (JR, NF and MV) collabora-
tively categorised all studies according to the type of control: active psy-
chological treatment (e.g. cognitive therapy, EMDR), psychological
placebo (e.g. stress management training), pharmacological treatment
(e.g. SSRI), wait-list or treatment as usual (TAU). There were no studies
that relied on a pill placebo control. Studies were also grouped and
analysed according to whether the study population comprised adults
or children.
3.2.2. Moderators
Potential treatment moderators were examined including
patient-related factors (age, duration of illness, OCD severity at baseline,
depression scores at baseline and end-point) and study-related factors
(duration of treatment, control group type, details of control treatment
within each arm, experience level of therapists delivering CBT, and in-
formation about concurrent medication),
3.3. Bias
Each study was assessed for risk of bias using the Cochrane risk of
bias tool version 2.0 (RoB2: Higgina Higgins et al. [35]) by two authors
(JR and MV) and any discrepancies resolved by discussion. The RoB2 as-
sesses bias that may arise across five domains: bias from randomisation,
4
Comprehensive Psychiatry 106 (2021) 152223
deviations from intended interventions, missing outcome data, out-
come measurement and bias in selection of the reported results.
3.4. Exploratory outcomes
Exploratory outcomes, which emerged during the stage of data col-
lection and were therefore not preregistered at PROSPERO, included
analysis of the moderating effect of treatment fidelity and the presence
of researcher allegiance on effect size.
3.4.1. Treatment fidelity
Based on the descriptions given within each study, an assessment of
treatment fidelity was made by an independent CBT expert (LD). This
involved assessing each of the components of ERP deemed to be essen-
tial e.g. the presence of response prevention, the exposure being
prolonged, graded and regular, the therapy being collaborative and
the level of experience of the therapist. Each component was given a
score of between zero (insufficient information was available to make
a decision) and five (awarded where the component appeared was at
a level consistent with recognised ‘best practice’) with a maximum
available score of 35 (Individual scores are included in Table 1).
3.4.2. Researcher allegiance
Researcher allegiance was assessed for all trials utilising the ‘re-
searcher allegiance assessment tool’ used by Turner et al. [36]; adapted
from Cuijpers et al. [37]) in a recent meta-analysis that examined psy-
chological interventions in psychosis. Following Turner et al. (2014)
[36], we posed the following questions to evaluate the presence of re-
searcher allegiance: Is only one of the interventions mentioned in the
title? In the introduction is one of the interventions explicitly described
as being the main experimental intervention? Was one intervention
specifically described as a control condition? Is there an explicit hypoth-
esis that one treatment is expected to be more effective than the other?
If the answer to any of these questions was yes, the study was deemed
at risk of researcher allegiance.
3.5. Analysis
Data were initially extracted independently by two of the authors
(JR and NF), and were then independently re-extracted by another au-
thor (KL), with differences being resolved.
Pooled effect sizes were calculated using Comprehensive Meta-
analysis, version 3. The effect size employed was Hedges g, which is
the standardised difference between means, corrected for the tendency
towards overestimation in small studies (Hedges, 1981). Effect sizes
were calculated comparing end-of trial total Y-BOCS (or alternative
scale) scores for the intervention and control groups. Random-effects
models were used in all analyses.
Heterogeneity was examined by use of Q and I2 statistics. An I2 value
of 0–40% suggests that heterogeneity may not be important, 30–60%
may represent moderate heterogeneity, 50–90% may represent sub-
stantial heterogeneity, and 75–100% may represent considerable het-
erogeneity (see Higgins & Green, Cochrane Handbook, 2011 [38]).
Publication bias was examined using the statistical techniques of
Duval and Tweedie's (2000) [39] trim and fill, which aims to estimate
the number of missing studies within an analysis and the effect that
those studies might have on outcomes.
4. Results
Following our search strategy as outlined above, thirty-six trials [11],
[40-74 were included in the final analysis Fig. 1.
The trials involved 2020 participants (1005 receiving CBT with a
substantive ERP component, and 1015 assigned to a control condition).
The comparator control conditions were active psychological treatment
(k = 8), psychological placebo (k = 10), a pharmacological treatment
J.E. Reid, K.R. Laws, L. Drummond et al. Comprehensive Psychiatry 106 (2021) 152223

Table 1
Characteristics of studies included in meta-analysis.

Study Control arm Sample Age Females Therapy Duration of Baseline Concurrent CBT
number (years) (%) Time illness YBOCS medication grading
allowed

Van Oppen et al. cognitive therapy 57 34.7 53 720 13.07 24.8 no 23

1995
Lindsay et al. 1997 anxiety management 18 11 67 900 10.95 26.57 yes 25
Freeston et al. 1997 wait-list 29 35.8 45 2430 9.4 23.5 yes 34
de Haan et al. 1998 clomipramine 22 13.7 50 no info 0.9 22.5 no info 29
Van Balkom et al. Cognitive therapy 38 33.8 45 720 11.3 25.0 No 24
1998
McLean et al. 2001 cognitive therapy 76 35 no info 1800 no info 21.84 yes 33
Cottraux et al. 2001 cognitive therapy 65 35.8 74.6 1200 13.48 28.5 yes 8
Greist et al. 2002 Systematic Relaxation Therapy 122 (*218) 39 42 660 22 25 yes 27
Volpato Cordioli wait-list 47 36.5 51 1440 21.1 25.7 yes 6
et al.
POTS et al. 2004 sertraline 56 (*112) 11.7 50 840 no info 24.8 no 26
Barrett et al. 2004 wait-list 48 (*77) 11.25 42.5 1260 no info 23.3 yes 21
O'Connor et al. 2005 inference-based approach 32 (*54) 38.3 64 no info no info 22.3 no 27
Whittal et al. 2005 cognitive behavioural therapy (without 71 35 not 720 13 22.5 yes 11
ERP) provided
Nakatani et al. 2005 autogenic training 18 (*28) 34 66.8 540 12.9 30.2 no 28
Asbahr et al. 2005 Sertraline 40 13.7 25 1080 4.8 26.3 No 27
Foa et al. 2005 clomipramine 65 (*122) 34.9 55.4 1800 16.9 25.5 no info 26
Sousa et al. 2006 Sertraline 56 38.5 77 1440 23.5 25.1 No 25
Anderson et al. 2007 wait-list 38 (*61) 33.2 64.3 600 12.4 24 Yes 28
Belloch et al. 2008 cognitive therapy 29 32 6.1 1800 6 25.6 Yes 11
Freeman et al. 2008 relaxation therapy 42 7.11 not 780 no info 22.36 Yes 23
provided
Khodarahimi et al. satiation therapy 40 (*60) 24.6 0 1080 no info 37.2 No 23
2009
Piacentini et al. 2011 psychoeducation and relaxation 71 12.3 63.4 1080 no info 24.9 No 30
Storch et al. 2011 wait-list 31 11.1 39 1080 no info 23.4 yes 30
Andersson et al. attention control supportive therapy 101 34 66 no info 18 21.1 yes 24
2012
Belotto-Silva et al. fluoxetine 159 34 55.1 1800 no info 25.88 yes 21
2012
Visser et al. 2014 inference-based approach 90 34.8 65.7 1080 15.9 26.02 yes 21
Herbst et al. 2014 wait-list 34 35.6 65 no info 14 20.12 yes 24
Vogel et al. 2014 wait-list 20 (*30) 34.8 65 no info no info 23.8 no info 29
Lewin et al. 2014 treatment as usual 31 5.81 29 720 no info 24.5 Yes 22
Mahoney et al. 2014 treatment as usual 67 39.1 59.5 no info no info 33.1 (DOCS Yes 20
scale)
Freeman et al. 2014 Family-based relaxation therapy 127 7.2 67 780 no info 25.5 Yes 25
Marsden et al. 2016 eye movement desensitisation 55 32 61.8 no info no info 25.82 no info 28
reprogramming
Fineberg et al. 2018 sertraline 31 (*49) 33.8 57 960 no info 26.8 Yes 29
Lenhard et al. 2017 wait-list 67 14.6 46 no info no info 22.5 Yes 4
Kyrios et al. 2018 progressive relaxation training 179 33.4 65.7 no info 13.7 21.94 yes 16
Kobayashi et al. treatment as usual 18 30.1 47.1 960 10.2 27.2 yes 16
2019

(k = 7), wait-list control (k = 8) and ‘treatment-as-usual’ (k = 3). All
studies used the change in OCD symptom score ratings as the primary
outcome; in 35/36 studies this was the Y-BOCS. One study utilised the
Dimensional Obsessive-Compulsive Scale (DOCS (Mahoney et al.
2014). Most studies (64%) permitted the use of concurrent psychotropic
medication; 25% did not allow concurrent medication and a further 11%
did not clearly report this information.
t-tests performed showed that intervention and controls did not dif-
fer in relation to age, duration of OCD symptoms, baseline Y-BOCS score
or depression symptomatology.
The primary analysis demonstrated a large positive effect for CBT
with ERP compared to all controls in the reduction of Y-BOCS scores
(g = 0.74: 95% CI = 0.51 to 0.97). The studies were heterogeneous
(Q (35) = 206.812, p < 0.001) with an I2 value of 83.08. (Fig. 2:.)
4.1. Publication bias
Visual inspection of a funnel plot and analysis using Duval and
Tweedie's trim and fill method did not suggest the presence of publica-
tion bias. (Fig. 3)
4.2. Moderator and exploratory analyses
Several moderator analyses were undertaken to evaluate factors that
might affect the observed treatment effect size:
4.3. Control type
The control groups varied considerably across trials: active treat-
ments as controls (k = 8), psychological placebos (k = 10), pharmaco-
logical treatment (k = 7), wait-list (k = 8) and treatment as usual (k =
3). Two out of 7 studies comparing ERP with a pharmacotherapy control
involved clomipramine and the rest involved SSRI. Adequate dosages of
medication were provided in all but one study (Sousa et al. [56]), in
which only a maximum of half the recommended daily dosage of SSRI
(Sertraline100mg/day) was provided in the control arm.
Subgroup analyses revealed a significant benefit for CBT with ERP
when compared to: psychological placebo (g = 1.13 95% CI 0.71 to
1.55) and wait-list (g = 1.27: 95% CI 0.79 to 1.75). By contrast, trials
using active psychological interventions as a control revealed no signif-
icant benefit from CBT with ERP (g = −0.05: 95% CI −0.27 to 0.16). In

5
J.E. Reid, K.R. Laws, L. Drummond et al.
Fig. 1. Studies included in systematic analysis.
the case of comparisons with trials of pharmacological treatment, when
all pharmacotherapy studies were included, a significant benefit for ERP
was seen (g; 0.36: 95% CI 0.07 to 0.64). However, when we excluded the
study utilising inadequate dosages of SSRI in the control arm, the effect
size when compared to pharmacological controls was only borderline
significant (g = 0.32: 95% CI -0.00 to 0.64, p = 0.05). Treatment as
usual trials were not analysed as only three such studies were located.
The difference in effect sizes between the groups was significant
(Q = 47.62, p < 0.001).
We calculated the end of trial Y-BOCS mean difference score for CBT
with ERP when compared to each intervention type. This was greatest
for treatment as usual (10.8: 95% CI: 7.17 to 14.42) and waitlist (8.53:
95% CI: 5.15 to 11.91) and smallest for CBT with ERP compared to active
psychological treatment (−0.38: 95% CI: −2.09 to 1.31). In comparison
to psychological placebo the mean difference was 5.87 (95% CI: 3.81 to
7.93) and when compared to pharmacological treatment it was 3.15
(95% CI: 0.68 to 5.62).
4.4. Researcher allegiance bias
We performed a separate analysis grouping studies according to the
presence (k = 28) and absence (k = 8) of suspected researcher
6
Comprehensive Psychiatry 106 (2021) 152223
allegiance (Fig. 4:). In the studies identified as having researcher alle-
giance, a very large favourable effect for CBT with ERP emerged (g =
0.95: 95% CI 0.69 to 1.2). By contrast, in trials where researcher alle-
giance was not identified (k = 8), CBT with ERP showed no significant
effect on Y-BOCS scores (g = 0.02: 95% CI −0.29 to 0.33). The difference
in effect sizes between the two groups was significant (Q = 20.33
p < 0.005).
The mean difference in Y-BOCS score at the end of trial between ERP
and controls in for studies with researcher allegiance was 5.99 (95% CI:
4.47 to 7.50) and 0.16 (95% CI: −2.45 to 2.77) for those studies without
researcher allegiance.
4.5. Adults vs children
Another sub-group analysis was completed to evaluate the effect
sizes in studies that were conducted in adult populations (k = 26) ver-
sus those involving children (k = 10). The effect size for CBT with ERP
was significant both for children (g = 1.09: 95% CI 0.60 to 1.58) and
for adults (g = 0.60: 95% CI 0.35 to 0.84). The difference between the
two groups was not significant (Q = 3.14, p = 0.08). The mean differ-
ence in Y-BOCS score between ERP and controls at the end of trial was
J.E. Reid, K.R. Laws, L. Drummond et al.
Fig. 2. Forest plot showing effect sizes for all included studies.
4.23 (95% CI: 2.42 to 6.03) for adults and 6.51 (95% CI: 3.88 to 9.14) for
children.
4.6. Group vs individual therapy
A planned analysis comparing trials of group versus individual
CBT was not completed due to too few studies (k = 5) employing
group CBT.
4.7. Other moderator analyses
The impact of age, proportion of females, duration of illness (mea-
sured in years), baseline severity (Y-BOCS scores), depression baseline
scores on the Beck Depression Inventory (BDI), change in depression
scores, total amount of therapy time (measured in minutes) and the
fidelity quality of ERP as continuous variables were assessed using
7
Comprehensive Psychiatry 106 (2021) 152223
meta-regression) Table 2. Only the age of the participant had a signifi-
cant inverse effect on the effect size.
4.8. Risk of bias
On the Cochrane Risk of Bias assessment tool (RoB2) Fig. 5 only 8
studies (22%) were deemed as being at low risk of bias; 14 studies
(39%) were assessed as being at high risk of bias and a further 14
(39%) were highlighted as having concerns about bias. In 56% of the
studies, potential concerns emerged about the randomisation process,
frequently owing to a lack of detail provided in the published trials.
An exploratory analysis (suggested by a reviewer) revealed no sig-
nificant difference (Q = 0.13, df = 2, p = 0.94) in effect sizes for studies
rated as at: high risk of bias g = 0.70, 0.53 to 0.98, k = 13, some concern
(g = 0.79, 0.50 to 1.08, k = 15) and low risk of bias (g = 0.72, 0.14 to
1.29 k = 8). Despite this, trials at low risk of bias had extremely wide
J.E. Reid, K.R. Laws, L. Drummond et al. Comprehensive Psychiatry 106 (2021) 152223

Fig. 3. Funnel plot.

Table 2 9 (25%) of studies explicitly stated they did not allow concurrent
Continuous variables as potential moderators of Y-BOCS effect size at end of trial. medication (Table 1), meaning that for the majority of studies, a pro-
Co-variate k Range Coefficient Z-Value p-value portion of patients were in reality, receiving combination therapy
Therapy Time (mins) k = 27 540–2430 −0.0002 −0.64 0.52
rather than CBT with ERP as monotherapy. This observation aligns
Baseline Y-BOCS score k = 36 19.2–37.2 0.0061 0.18 0.86 with the findings of Skapinakis et al. (2016 [16]). Although the num-
Depression baseline score k = 15 8.9–24.9 0.0415 0.85 0.40 ber of studies utilising clearly defined CBT with ERP monotherapy is
(BDI) small (k = 9), we noted that there were differences between these
Duration of OCD symptoms k = 20 0.9–23.5 0.0005 −0.02 0.99
groups; the participants in the monotherapy studies were younger,
(years)
Reduction in depressive k = 22 16.4–78.8 −0.0078 −0.96 0.34 had higher baseline Y-BOCS scores and also received less therapy
symptoms time than in the other studies. Interestingly, the studies of CBT with
Female (%) k = 33 0–83.0 −0.0009 −0.11 0.91 ERP monotherapy had numerically higher CBT grading scores
Mean Age k = 36 5.76–39.0 −0.0272 −2.73 0.006 (Table 1). This could indicate that the monotherapy studies in this
CBT grading k = 36 4.00–34.0 0.0217 1.35 0.17
meta-analysis had greater CBT fidelity and/or more rigorous reporting
of their methodology.
This meta-analysis therefore adds weight to current recommenda-
confidence intervals (0.14 to 1.29), suggesting imprecision and that fur- tions regarding the effectiveness of ERP, but does not indicate a signifi-
ther low risk trials are required. cant superiority for ERP compared to other active psychological
treatments including some currently posited to be effective for OCD
and raises questions about the superiority of ERP over pharmacological
5. Discussion treatment for OCD. Consequently, our findings question current clinical
guidance (e.g. the 2005 NICE guidelines [8]) prioritising the use of CBT
This comprehensive meta-analysis, involving 36 RCTs and 2020 with ERP over other CBT modalities or pharmacological treatment for
participants, evaluates the effectiveness of CBT with ERP in reducing those with OCD. Indeed, NICE [8] acknowledges that alongside ERP,
OCD symptoms. Our analysis demonstrated a large pooled effect clinicians working in the field of OCD frequently provide “different var-
size for CBT with ERP (g = 0.74: 95% CI = 0.51 to 0.97). As far as iants” of cognitive therapy as well as a combination of cognitive therapy
we are aware, this is the first meta-analysis performed in recent and ERP applied as a “coherent package’.
years to focus exclusively on CBT with ERP as the investigational Furthermore, NICE [8] asserts that whereas “ERP and cognitive ther-
treatment. apy have different theoretical underpinnings”, as, for example, “most
Our analyses revealed that the effect size attributed to CBT with ERP current cognitive therapy explicitly seeks behaviour change but is not
depends strongly on the choice of control comparator. Thus, CBT with operating within a habituation paradigm”, it is “uncertain whether ei-
ERP was more efficacious than psychological placebo (such as ther treatment is superior to the other, or indeed whether combining
psychoeducation, progressive relaxation therapy and autogenic train- these interventions confers any added benefit (Abramowitz [12]).
ing), with a large pooled effect size (g = 1.13 95% CI 0.71 to 1.55). By NICE [8] additionally draws attention to the difficulty of comparing
contrast, when compared to active psychological treatments such as ERP with different variants of CBT owing to a tendency for the treatment
cognitive therapy or EMDR, the pooled effect size of CBT with ERP was modalities to overlap.
not significant (g = 0.05: 95% CI -0.27 to 0.16, p = 0.62). Similarly, Turning to risk of bias, concerns were identified for three-quarters of
whereas our analysis found a small advantage for CBT with ERP when all published ERP trials (78%), highlighting the need for studies with ro-
compared to pharmacological treatment across all available trials bust methodology and rigorous reporting. Aside from the far fewer
(g = 0.36: 95% CI 0.07 to 0.64), the advantage is marginal when ade- numbers of trials at low risk of bias, they also had very wide confidence
quate dosages of pharmacological treatment are provided in the control intervals, suggesting high imprecision in the effect size estimate and
arm (g = 0.32: 95% CI -0.00 to 0.64, p = 0.05). that further low risk trials are required. After 25 years of ERP for OCD tri-
In terms of interpreting the findings, it is notable that most ERP als, currently fewer than 300 participants have received ERP for OCD in
trials have permitted concurrent medication at stable doses; only existing high quality trials.

8
J.E. Reid, K.R. Laws, L. Drummond et al.
Fig. 4. Forest plot for studies grouped by presence of researcher allegiance.
A driver of the overall confidence interval is, of course, the individual
trial sample sizes that contribute to the pooled effect. As noted, trials at
low risk of bias had the widest confidence intervals, although those for
high risk trials were also wide. We note that high risk trials had – on av-
erage – 50% fewer participants than low risk trials and are insufficiently
powered to detect even the large effect size reported here. By contrast,
low risk trials were sufficiently powered to detect the large effect size.
Thus, while the wide confidence intervals for high risk trials may reflect
underpowering, the wider confidence intervals for low risk trials would
not seem to be wholly attributable to the under-powering of trials.
In addition to assessing risk of bias, this is the first study to look at
the role of researcher allegiance in studies of CBT with ERP in the treat-
ment of OCD. Our results demonstrate a clear difference in effect size re-
lated to our ratings of researcher allegiance. Notably, over three quarters
of all trials (28/36: 78%) were classified as showing researcher alle-
giance and those trials produced a large significant effect size (g =
0.95: 95% CI 0.69 to 1.2). Indeed, all studies that demonstrated a statis-
tically significant beneficial effect for CBT with ERP were assessed as
having researcher allegiance. By contrast, in studies not displaying re-
searcher allegiance, the effect of CBT with ERP was nonsignificant
(g = 0.02: 95% CI -0.29 to 0.33, p 0.89). The number of trials without re-
searcher allegiance was small (k = 8) and so some caution may be re-
quired when interpreting this result. Nevertheless, the finding that
78% of studies included in this meta-analysis were evaluated as being
at high risk of researcher allegiance is concerning in itself and highlights
this as an area of research that requires further exploration.
9
Comprehensive Psychiatry 106 (2021) 152223
Researcher allegiance appears to operate, in part, via study design
features favouring not only the preferred treatment, but crucially per-
haps the preferred control comparison. In this context, we note that
the trials with high and low researcher allegiance differed markedly in
terms of the types of control group they employed.1 While 7/8 (87.5%)
low-allegiance trials employed active controls, only 8/28 (28.6%) high
allegiance trials had active controls. In fact, only high-allegiance trials
used TAU or wait list control (wait-list comprising almost 30% of all
high-allegiance trials). That wait-list controls can inflate effect sizes in
psychotherapy trials is well-known and one possibility is by creating
nocebo effects in control groups (Furukawa et al.) [30]. For the corpus
of trials examined in the current meta-analysis, the average pre-post
change in Y-BOCS scores for wait-list arms was just 3.3%, compared
with 30.9% for all other control comparators combined (and 22.6%
after removing ‘active’ control arms involving CBT). In their discussion
of control choice, Leichsenring & Steinert (2017) [75] remarked cor-
rectly that “When examining efficacy, a treatment may be compared
with different comparators, that is, with an established treatment, treat-
ment as usual, a placebo, or a waiting list, with decreasing strictness of the
empirical test.” (p.1323, our italics). The adoption of a wait-list control
condition in psychotherapy trials is not only a possible methodological
weakness, but a mechanism that allows researcher allegiance to poten-
tially interfere (Dragioti et al. [76]).
1
We thank a reviewer for noting this
J.E. Reid, K.R. Laws, L. Drummond et al.
Fig. 5. ROB2 assessment.
Moreover, the effects of researcher allegiance persist even after
methodological quality is controlled (Munder et al. [77]). Researcher al-
legiance thus apparently exerts effects beyond designing the study in a
way which benefits the preferred treatment. This does not imply nefar-
ious intent of researchers; it may simply reflect that researcher enthusi-
asm or expertise for a preferred treatment is not fully represented in the
variables commonly coded as methodological characteristics. However,
one potential clinical implication is that the efficacy of CBT with ERP is
dependent upon therapist factors and, importantly, that patients may
have an inferior outcome if treated in centres whose expertise does
not focus on ERP.
We also evaluated the fidelity of the CBT with ERP within the indi-
vidual studies. In terms of statistical analysis, we could not demonstrate
that the quality of CBT with ERP (as graded by an independent assessor)
impacted upon its effectiveness. However, during the process of
10
Comprehensive Psychiatry 106 (2021) 152223
extracting the data, it was noted that sufficiently detailed information
about study design, in particular with regards to the fidelity of the CBT
with ERP delivered, was frequently missing. This would be expected to
have impacted upon the reliability of the grading in our analysis.
Our findings therefore concur with the conclusions of previous
meta-analyses (Ӧst et al. [28], Skapinakis et al. [16]), which highlighted
problems with methodological quality in published studies on CBT for
OCD. These inadequacies substantially compromise interpretation of
the existing evidence. As a result, caution should be applied when
attempting to draw conclusions and guide treatment based on results
from such studies. A pressing need exists for high quality studies of
CBT with ERP in order to establish clinical utility.
Moderator analyses typically address questions about who might
most benefit from an intervention (such as CBT) and how we might im-
prove outcomes for those who benefit least. We found an inverse
J.E. Reid, K.R. Laws, L. Drummond et al.
Fig. 5 (continued).
relationship between effect size and age on meta-regression, suggesting
ERP may work best in the young. Results from naturalistic longitudinal
studies suggest younger patients may respond more favourably to treat-
ment (Mancebo et al. 2014) [78]. It is possible that this result however
reflects a confound with duration of illness, which was found not to sig-
nificantly affect the ES in this analysis, but as the sample giving duration
of illness data was small this analysis was probably underpowered.
Aside from age, we found a series of null relationships for the mod-
erators examined (therapy time, baseline Y-BOCS severity, baseline de-
pression, reduction of depression symptoms, OCD duration, proportion
of female participants, and CBT with ERP fidelity rating). The failure to
find significant (sample, treatment, or participant) predictors for effi-
cacy accords with the meta-analytic findings of Olatunji et al. (2013)
[29], who also failed to detect any significant impact of plausible
moderators on CBT outcomes. Although such findings may appear
counter-intuitive, they are consistent with the notion that CBT may be
beneficial to most people diagnosed with OCD: As noted by Olatunji
et al. (2013) [29], the absence of significant moderators potentially
“highlights the effectiveness of CBT for patients with wide range of
symptom complexity”. (p. 39).
However, variables such as ‘baseline severity’ and ‘amount of treat-
ment delivered’ would reasonably be expected to affect the outcome
of an efficacious treatment. We therefore suggest that the demonstrable
lack of methodological rigor and reporting could also explain why, as of
11
Comprehensive Psychiatry 106 (2021) 152223
yet, we lack reliable evidence to indicate which patient and treatment
factors might predict response and thus guide clinical decision-
making. It is possible that other study and patient characteristics have
a moderating effect, but as yet remain unstudied. For example, ‘adher-
ence to homework tasks’ has previously been demonstrated to be a sig-
nificant predictor of treatment outcome (Simpson et al. [79]) but is not
consistently reported in studies.
A change of 5 Y-BOCS points is known to represent a clinically
meaningful change in clinical status (Hollander [1]. We may therefore
interpret the mean between-intervention difference in the end-of-trial
Y-BOCS scores as an additional crude estimate of the ‘clinical impor-
tance’ of the reported effect sizes. (This metric assumes the comparator
groups had equivalent scores at baseline.) Using 5 Y-BOCS points as a
benchmark, it can be seen that CBT with ERP produced clinically impor-
tant advantages compared to treatment as usual, waitlist and psycho-
logical placebo. In contrast, there was no clinically important
advantage for ERP over other active forms of psychological treatment
and the advantage over pharmacological treatment was indeterminate.
Furthermore, based on Y-BOCS differences, the effect of ERP appeared
clinically important only in those studies where researcher allegiance
was present and appeared more robust in studies of children than
adults.
Finally, discrepancies emerge between the large effect size demon-
strated for CBT with ERP in this analysis (which is supported by the ef-
fect sizes seen in previous analyses) and the more modest effects
observed in naturalistic clinical practice, in which setting a sizeable pro-
portion of patients fail to achieve benefit (Eisen et al. [80]). This could
potentially be due to the existence of moderating variables which, to
date, remain unidentified. Uncovering such moderators may enable us
to predict for which patients and in which conditions CBT with ERP is
likely to be most effective. Future studies of CBT with ERP should be de-
signed with sufficient statistical power to enable the identification of
such moderators.
6. Conclusions
Having subjected all the published randomised controlled trials of
CBT with ERP for OCD to meta-analysis, we find this intervention ap-
pears to be effective. However, when the studies were parsed according
to the choice of comparator control, no advantage for CBT with ERP was
found in those studies where an active psychological treatment was
used as the control comparator. This casts doubt on the superiority of
CBT with ERP over other forms of psychological therapy for OCD.
Whereas CBT with ERP was significantly superior to pharmacological
treatment when all eligible studies were analysed, the effect became
marginal and only approached significance when adequate pharmaco-
therapy dosages were used in the control arm. Our meta-analysis fur-
ther highlights concerns about the methodological rigor and reporting
of published studies of CBT with ERP in OCD, with only a minority of
studies deemed to be at low risk of bias. In addition, an exploratory anal-
ysis revealed that the positive effect for CBT with ERP was restricted to
those studies showing evidence of researcher allegiance in favour of
CBT with ERP, questioning its generalizability. This finding highlights
the need for further research into the presence and implications of re-
searcher allegiance within studies of psychological therapy in OCD. In
sum, an unmet need remains for rigorously designed randomised con-
trolled trials to investigate the patient and treatment related factors
governing the efficacy of CBT with ERP for OCD.
Declaration of Competing Interest
Prof. Naomi A. Fineberg declares that in the past 3 years she has held
research or networking grants from the ECNP, UK NIHR, EU H2020,
MRC, University of Hertfordshire; she has accepted travel and/or hospi-
tality expenses from the BAP, ECNP, RCPsych, CINP, International Forum
of Mood and Anxiety Disorders, World Psychiatric Association, Indian
J.E. Reid, K.R. Laws, L. Drummond et al.
Association for Biological Psychiatry, Sun; she has received payment
from Taylor and Francis and Elsevier for editorial duties. In the past
3 years, she has accepted a paid speaking engagement in a webinar
sponsored by Abbott. Previously, she has accepted paid speaking en-
gagements in various industry supported symposia and has recruited
patients for various industry-sponsored studies in the field of OCD treat-
ment. She leads an NHS treatment service for OCD. She holds Board
membership for various registered charities linked to OCD. She gives ex-
pert advice on psychopharmacology to the UK MHRA.
Dr. Jemma Reid, Prof. Keith Laws, Dr. Matteo Vismara and Dr.
Benedetta Grancini report no financial relationships with commercial
interests.
References
[1] Hollander E, Stein DJ, Fineberg NA, Marteau F, Legault M. Quality of life outcomes in
patients with obsessive-compulsive disorder: relationship to treatment response
and symptom relapse. J Clin Psychiatry. 2010;71(6):784–92.
[2] APA. Diagnostic and Statistical Manual of Mental Disorders. 5th ed.. American Psy-
chiatric Association; 2013.
[3] Fineberg NA, Hengartner MP, Bergbaum C, Gale T, Rössler W, Angst J. Lifetime co-
morbidity of obsessive-compulsive disorder and sub-threshold obsessive-
compulsive symptomatology in the community: impact, prevalence, socio-
demographic and clinical characteristics. Int J Psychiatry Clin Pract. 2013;17(3):
188–96.
[4] Dell’Osso B, Benatti B, Grancini B, Vismara M, De Carlo V, Cirnigliaro G, et al. Investi-
gating duration of illness and duration of untreated illness in obsessive compulsive
disorder reveals patients remain at length pharmacologically untreated. Int J Psychi-
atry Clin Pract. 2019;23(4):311–3.
[5] Fineberg NA, Dell’Osso B, Albert U, et al. Early intervention for obsessive compulsive
disorder: an expert consensus statement. Eur Neuropsychopharmacol. 2019;29(4):
549–65.
[6] Fineberg NA, Drummond LM, Reid J, Cinosi E, Carmi L, Mpavaenda DN, et al. Manage-
ment and treatment of obsessive–compulsive disorder. New Oxford Textbook of
Psychiatry; 2020.
[7] Koran LM, Hanna GL, Hollander E, Nestadt GSH. Practice guideline for the treatment
of patients with obsessive-compulsive disorder. Am J Psychiatry. 2007;164:5–53.
[8] NICE. Obsessive-compulsive disorder and body dysmorphic disorder: treatment.
Nice (National Inst Heal Care Excell.); 2005.
[9] NICE. 2019 surveillance of obsessive-compulsive disorder and body dysmorphic dis-
order: treatment (NICE guideline CG31) [Internet]. [cited 2020 Jun 22]. Available
from: https://www.nice.org.uk/guidance/cg31/resources/2019-surveillance-of-
obsessivecompulsive-disorder-and-body-dysmorphic-disorder-treatment-nice-
guideline-cg31-6713804845/chapter/Surveillance-decision?tab=evidence; 2019.
[10] Baldwin DS, Anderson IM, Nutt DJ, Allgulander C, Bandelow B, Den Boer JA, et al.
Evidence-based pharmacological treatment of anxiety disorders, post-traumatic
stress disorder and obsessive-compulsive disorder: a revision of the 2005 guidelines
from the British Association for Psychopharmacology. J Psychopharmacol. 2014;28
(5):403–39.
[11] Fineberg N, Baldwin D, Drummond L, Wyatt S, Hanson J, Gopi S, et al. Optimal
Tretament for OCD (OTO): a randomised controlled feasibility trial comparing the
clinical and cost effectiveness of cognitive Behavioural therapy (CBT) and selective
serotonin reuptake inhibitors (SSRI) and their combination in the Management of
Obsessive Compulsive Disorder. Eur Neuropsychopharmacol. 2018;28(6):779.
[12] Uhre CF, Uhre VF, Lonfeldt NN, Pretzmann L, Vangkilde S, Plessen KJ, et al. Systematic
review and meta-analysis: cognitive-behavioral therapy for obsessive-compulsive
disorder in children and adolescents. J Am Acad Child Adolesc Psychiatry. 2019;59
(1):64–77.
[13] Storch EA, Peris TS, De Nadai A, Piacentini J, Bloch M, Cervin M, et al. Little doubt that
CBT works for pediatric OCD. J Am Acad Child Adolesc Psychiatry. 2020;59(7):
785–7.
[14] Geller DA, Hosker D. When science challenges our long-held assumptions about the
robustness of evidence for standard of care. J Am Acad Child Adolesc Psychiatry.
2020;59(7):792–3.
[15] Uhre VF, Uhre CF, Lønfeldt NN, Pretzmann L, Vangkilde S, Plessen KJ, et al. Dr. Uhre
et al. Reply: Journal of the American Academy of Child and Adolescent Psychiatry;
2020.
[16] Skapinakis P, Caldwell D, Hollingworth W, Bryden P, Fineberg N, Salkovskis P, et al. A
systematic review of the clinical effectiveness and cost-effectiveness of pharmaco-
logical and psychological interventions for the management of obsessive–
compulsive disorder in children/adolescents and adults. Health Technol Assess.
2016;20(43):1–392.
[17] Skapinakis P, Caldwell DM, Hollingworth W, Bryden P, Fineberg NA, Salkovskis P,
et al. Pharmacological and psychotherapeutic interventions for management of
obsessive-compulsive disorder in adults: a systematic review and network meta-
analysis. Lancet Psychiatry. 2016;3(8):730–9.
[18] Skapinakis P, Caldwell DM, Hollingworth W, Welton NJ, Fineberg N, Salkovskis P,
et al. Network meta-analyses and treatment recommendations for obsessive-
compulsive disorder – Authors’ reply. Lancet Psychiatry. 2016;3(10):921–2.
[19] Christensen H, Hadzi-Pavlovic D, Andrews G, Mattick R. Behavior therapy and tricy-
clic medication in the treatment of obsessive-compulsive disorder: a quantitative re-
view. J Consult Clin Psychol. 1987;55(5):701–11.
12
Comprehensive Psychiatry 106 (2021) 152223
[20] Abramowitz JS. Variants of exposure and response prevention in the treatment of
obsessive-compulsive disorder: a meta-analysis Behav Ther. Autumn. 1996;27(4):
583–600.
[21] Cuijpers P, Weitz E, Cristea IA, Twisk J. Pre-post effect sizes should be avoided in
meta-analyses. Epidemiol Psychiatr Sci. 2017;26(4):364–8.
[22] Abramowitz JS. Effectiveness of psychological and pharmacological treatments for
obsessive-compulsive disorder: a quantitative review. J Consult Clin Psychol. 1997;
65(1):44–52.
[23] Kobak KA, Greist JH, Jefferson JW, Katzelnick DJ, Henk HJ. Behavioral versus pharma-
cological treatments of obsessive compulsive disorder: a meta-analysis. Psychophar-
macology. 1998;136(3):205–16.
[24] Eddy KT, Dutra L, Bradley R, Westen D. A multidimensional meta-analysis of psycho-
therapy and pharmacotherapy for obsessive-compulsive disorder. Clin Psychol Rev.
2004;24(8):1011–30.
[25] Rosa-Alcázar AI, Sánchez-Meca J, Gómez-Conesa A, Marín-Martínez F. Psychological
treatment of obsessive-compulsive disorder: a meta-analysis. Clin Psychol Rev.
2008;28(8):1310–25.
[26] McGuire JF, Piacentini J, Lewin AB, Brennan EA, Murphy TK, Storch EA. A meta-
analysis of cognitive behavior therapy and medication for child obsessive-
compulsive disorder: moderators of treatment efficacy, response, and remission.
Depress Anxiety. 2015;32(8):580–93.
[27] Öst LG, Riise EN, Wergeland GJ, Hansen B, Kvale G. Cognitive behavioral and phar-
macological treatments of OCD in children: a systematic review and meta-analysis.
J Anxiety Disord. 2016;43:58–69.
[28] Öst LG, Havnen A, Hansen B, Kvale G. Cognitive behavioral treatments of obsessive-
compulsive disorder. A systematic review and meta-analysis of studies published
1993-2014. Clin Psychol Rev. 2015;40:156–69.
[29] Olatunji BO, Davis ML, Powers MB, Smits JAJ. Cognitive-behavioral therapy for
obsessive-compulsive disorder: a meta-analysis of treatment outcome and modera-
tors. J Psychiatr Res. 2013;47(1):33–41.
[30] Furukawa TA, Noma H, Caldwell DM, Honyashiki M, Shinohara K, Imai H, et al.
Waiting list may be a nocebo condition in psychotherapy trials: a contribution
from network meta-analysis. Acta Psychiatr Scand. 2014;130(3):181–92.
[31] Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to meta-analysis;
2009.
[32] Hezel D, Simpson H. Exposure and response prevention for obsessive-compulsive
disorder: a review and new directions. Indian J Psychiatry. 2019;61(7):85–92.
[33] Leykin Y, DeRubeis RJ. Allegiance in psychotherapy outcome research: Separating
association from bias. Clin Psychol Sci Pract. 2009;16(1):54–65.
[34] Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. PRISMA-P
statement 2015. Syst Rev. 2015;62(10):e1-34.
[35] Higgins JPT, Savović J, Page MJ, Elbers RGSJ. Chapter 8: Assessing risk of bias in a ran-
domized trial|Cochrane Training. Cochrane Handbook for Systematic Reviews of In-
terventions version 60 (updated July 2019); 2019.
[36] Turner DT, Van Der Gaag M, Karyotaki E, Cuijpers P. Psychological interventions for
psychosis: a meta-analysis of comparative outcome studies. Am J Psychiatry. 2014;
171(5):523–38.
[37] Cuijpers P, Driessen E, Hollon SD, van Oppen P, Barth J, Andersson G. The efficacy of
non-directive supportive therapy for adult depression: a meta-analysis. Clin Psychol
Rev. 2012;32(4):280–91.
[38] Higgins JPT GS. Cochrane handbook for systematic reviews of interventions version
5.1.0. The cochrane collaboration; 2011.
[39] Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and
adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455–63.
[40] Van Oppen P, De Haan E, Van Balkom AJLM, Spinhoven P, Hoogduin K, Van Dyck R.
Cognitive therapy and exposure in vivo in the treatment of obsessive compulsive
disorder. Behav Res Ther. 1995;
[41] Lindsay M, Crino R, Andrews G. Controlled trial of exposure and response prevention
in obsessive- compulsive disorder. Br J Psychiatry. 1997;
[42] . Freeston MH, Ladouceur R, Gagnon F, Thibodeau N, Rhéaume J, Letarte H, et al.
Cognitive-behavioral treatment of obsessive thoughts: A controlled study. J Consult
Clin Psychol. 1997;
[43] De Haan E, Hoogduin KAL, Buitelaar JK, Keijsers GPJ. Behavior therapy versus clo-
mipramine for the treatment of obsessive- compulsive disorder in children and ad-
olescents. J Am Acad Child Adolesc Psychiatry. 1998;
[44] Van Balkom AJLM, De Haan E, Van Oppen P, Spinhoven P, Hoogduin KAL, Van Dyck
R. Cognitive and behavioral therapies alone versus in combination with fluvoxamine
in the treatment of obsessive compulsive disorder. J Nerv Ment Dis. 1998;
[45] McLean PD, Whittal ML, Thordarson DS, Taylor S, Söchting I, Koch WJ, et al. Cognitive
versus behavior therapy in the group treatment of obsessive-compulsive disorder. J
Consult Clin Psychol. 2001;
[46] Cottraux J, Note I, Yao SN, Lafont S, Note B, Mollard E, et al. A randomized controlled
trial of cognitive therapy versus intensive behavior therapy in obsessive compulsive
disorder. Psychother Psychosom. 2001;
[47] Greist et al. Behaviour Therapy for Obsessive-Compulsive Disorder Guided by a
Computer or by a Clinician compared with relaxation as a control. J Clin Psychiatry
2002;63(2):138-145
[48] Cordioli AV, Heldt E, Bochi DB, Margis R, De Sousa MB, Tonello JF, et al. Cognitive-
behavioral group therapy in obsessive-compulsive disorder: A randomized clinical
trial. Psychother Psychosom. 2003;
[49] March JS. Cognitive-behavior therapy, sertraline, and their combination for children
and adolescents with obsessive-compulsive disorder: The pediatric OCD treatment
study (POTS) randomized controlled trial. J Am Med Assoc. 2004;
[50] Barrett P, Healy-Farrell L, March JS. Cognitive-behavioral family treatment of child-
hood obsessive-compulsive disorder: A controlled trial. J Am Acad Child Adolesc
Psychiatry. 2004;
J.E. Reid, K.R. Laws, L. Drummond et al.
[51] Ramos Asbahr F, Castillo AR, Montenegro Ito L, Dias De Oliveira Latorre MDR, Nunes
Moreira M, Lotufo-Neto F. Group cognitive-behavioral therapy versus sertraline for
the treatment of children and adolescents with obsessive-compulsive disorder. J
Am Acad Child Adolesc Psychiatry. 2005;
[52] O’Connor KP, Aardema F, Bouthillier D, Fournier S, Guay S, Robillard S, et al. Evalua-
tion of an inference-based approach to treating obsessive-compulsive disorder. Cogn
Behav Ther. 2005;
[53] Whittal ML, Thordarson DS, McLean PD. Treatment of obsessive-compulsive disor-
der: Cognitive behavior therapy vs. exposure and response prevention. Behav Res
Ther. 2005;
[54] Nakatani E, Nakagawa A, Nakao T, Yoshizato C, Nabeyama M, Kudo A, et al. A ran-
domized controlled trial of Japanese patients with obsessive-compulsive disorder -
Effectiveness of behavior therapy and fluvoxamine. Psychother Psychosom. 2005;
[55] Foa EB, Liebowitz MR, Kozak MJ, Davies S, Campeas R, Franklin ME, et al. Random-
ized, placebo-controlled trial of exposure and ritual prevention, clomipramine, and
their combination in the treatment of obsessive-compulsive disorder. Am J Psychia-
try. 2005;
[56] Sousa MB, Isolan LR, Oliveira RR, Manfro GG, Cordioli A V. A randomized clinical trial
of cognitive-behavioral group therapy and sertraline in the treatment of obsessive-
compulsive disorder. J Clin Psychiatry. 2006;
[57] Anderson RA, Rees CS. Group versus individual cognitive-behavioural treatment for
obsessive-compulsive disorder: A controlled trial. Behav Res Ther. 2007;
[58] Belloch A, Cabedo E, Carrió C. Empirically grounded clinical interventions: Cognitive
versus behaviour therapy in the individual treatment of obsessive-compulsive disor-
der: Changes in cognitions and clinically significant outcomes at post-treatment and
one-year follow-up. Behav Cogn Psychother. 2008;
[59] Freeman JB, Garcia AM, Coyne L, Ale C, Przeworski A, Himle M, et al. Early childhood
OCD: Preliminary findings from a family-based cognitive-behavioral approach. J Am
Acad Child Adolesc Psychiatry. 2008;
[60] Khodarahimi S. Satiation therapy and exposure response prevention in the treat-
ment of obsessive compulsive disorder. J Contemp Psychother. 2009;
[61] Piacentini J, Bergman RL, Chang S, Langley A, Peris T, Wood JJ, et al. Controlled com-
parison of family cognitive behavioral therapy and psychoeducation/relaxation
training for child obsessive-compulsive disorder. J Am Acad Child Adolesc Psychia-
try. 2011;
[62] E.A. S, N.E. C, J.R. M, A.B. L, A. R, L. B, et al. Preliminary investigation of web-camera
delivered cognitive-behavioral therapy for youth with obsessive-compulsive disor-
der. Psychiatry Research. 2011.
[63] Andersson E, Enander J, Andrén P, Hedman E, Ljótsson B, Hursti T, et al. Internet-
based cognitive behaviour therapy for obsessive-compulsive disorder: A random-
ized controlled trial. Psychol Med. 2012;
[64] Belotto-Silva C, Diniz JB, Malavazzi DM, Valério C, Fossaluza V, Borcato S, et al. Group
cognitive-behavioral therapy versus selective serotonin reuptake inhibitors for
obsessive-compulsive disorder: A practical clinical trial. J Anxiety Disord. 2012;
[65] Visser HA, Van Megen H, Van Oppen P, Eikelenboom M, Hoogendorn AW,
Kaarsemaker M, et al. Inference-Based Approach versus Cognitive Behavioral Ther-
apy in the Treatment of Obsessive-Compulsive Disorder with Poor Insight: A 24-
Session Randomized Controlled Trial. Psychother Psychosom. 2015;
[66] Herbst N, Voderholzer U, Thiel N, Schaub R, Knaevelsrud C, Stracke S, et al. No
talking, just writing! efficacy of an internet-based cognitive behavioral therapy
13
Comprehensive Psychiatry 106 (2021) 152223
with exposure and response prevention in obsessive compulsive disorder.
Psychother Psychosom. 2014;
[67] P.A. V, S. S, K. H, E.M. M, G. L, T.T. H, et al. A pilot randomized controlled trial of
videoconference-assisted treatment for obsessive-compulsive disorder. Behaviour
Research and Therapy. 2014.
[68] Lewin AB, Park JM, Jones AM, Crawford EA, DeNadai AS, Menzel J, et al. Family-based
exposure and response prevention therapy for preschool-aged children with
obsessive-compulsive disorder: A pilot randomized controlled trial. Behav Res
Ther. 2014;
[69] Freeman J, Sapyta J, Garcia A, Compton S, Khanna M, Flessner C, et al. Family-based
treatment of early childhood obsessive-compulsive disorder: The pediatric
obsessive-compulsive disorder treatment study for young children (POTS Jr) - A ran-
domized clinical trial. JAMA Psychiatry. 2014;
[70] Mahoney AEJ, Mackenzie A, Williams AD, Smith J, Andrews G. Internet cognitive be-
havioural treatment for obsessive compulsive disorder: A randomised controlled
trial. Behav Res Ther. 2014;
[71] Marsden Z, Lovell K, Blore D, Ali S, Delgadillo J. A randomized controlled trial com-
paring EMDR and CBT for obsessive–compulsive disorder. Clin Psychol Psychother.
2018;
[72] Lenhard F, Andersson E, Mataix-Cols D, Rück C, Vigerland S, Högström J, et al.
Therapist-Guided, Internet-Delivered Cognitive-Behavioral Therapy for Adolescents
With Obsessive-Compulsive Disorder: A Randomized Controlled Trial. J Am Acad
Child Adolesc Psychiatry. 2017;
[73] Kyrios M, Ahern C, Fassnacht DB, Nedeljkovic M, Moulding R, Meyer D. Therapist-
assisted internet-based cognitive behavioral therapy versus progressive relaxation
in obsessive-compulsive disorder: Randomized controlled trial. J Med Internet Res.
2018;
[74] Kobayashi Y, Kanie A, Nakagawa A, Takebayashi Y, Shinmei I, Nakayama N, et al. An
Evaluation of Family-Based Treatment for OCD in Japan: A Pilot Randomized Con-
trolled Trial. Front Psychiatry. 2020;
[75] Leichsenring F, Abbass A, Hilsenroth MJ, Leweke F, Luyten P, Keefe JR, et al. Biases in
research: risk factors for non-replicability in psychotherapy and pharmacotherapy
research. Psychol Med. 2017;47(6):1000–11.
[76] Dragioti E, Dimoliatis I, Fountoulakis KN, Evangelou E. A systematic appraisal of alle-
giance effect in randomized controlled trials of psychotherapy. Ann General Psychi-
atry. 2015;14:25.
[77] Munder T, Gerger H, Trelle S, Barth J. Testing the allegiance bias hypothesis: a meta-
analysis. Psychother Res. 2011;21(6):670–84.
[78] Mancebo MC, Boisseau CL, Garnaat SL, Eisen JL, Greenberg BD, Sibrava NJ, et al. Long-
term course of pediatric obsessive-compulsive disorder: 3 years of prospective
follow-up. Compr Psychiatry. 2014;55(7):1498–504.
[79] Simpson HB, Maher MJ, Wang Y, Bao Y, Foa EB, Franklin M. Patient adherence pre-
dicts outcome from cognitive behavioral therapy in obsessive-compulsive disorder.
J Consult Clin Psychol. 2011;79(2):247–52.
[80] Eisen JL, Sibrava NJ, Boisseau CL, Mancebo MC, Stout RL, Pinto A, et al. Five-year
course of obsessive-compulsive disorder: predictors of remission and relapse. J
Clin Psychiatry. 2013;74(3):233–9.